1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License, Version 1.0 only 6 * (the "License"). You may not use this file except in compliance 7 * with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 23 /* 24 * Copyright 2005 Sun Microsystems, Inc. All rights reserved. 25 * Use is subject to license terms. 26 */ 27 28 #pragma ident "%Z%%M% %I% %E% SMI" 29 30 #include <cma.h> 31 32 #include <strings.h> 33 #include <errno.h> 34 #include <time.h> 35 #include <fm/fmd_api.h> 36 #include <sys/fm/protocol.h> 37 38 cma_t cma; 39 40 cma_stats_t cma_stats = { 41 { "cpu_flts", FMD_TYPE_UINT64, "cpu faults resolved" }, 42 { "cpu_fails", FMD_TYPE_UINT64, "cpu faults unresolveable" }, 43 { "cpu_blfails", FMD_TYPE_UINT64, "failed cpu blacklists" }, 44 { "cpu_supp", FMD_TYPE_UINT64, "cpu offlines suppressed" }, 45 { "cpu_blsupp", FMD_TYPE_UINT64, "cpu blacklists suppressed" }, 46 { "page_flts", FMD_TYPE_UINT64, "page faults resolved" }, 47 { "page_fails", FMD_TYPE_UINT64, "page faults unresolveable" }, 48 { "page_supp", FMD_TYPE_UINT64, "page retires suppressed" }, 49 { "page_nonent", FMD_TYPE_UINT64, "retires for non-existent fmris" }, 50 { "bad_flts", FMD_TYPE_UINT64, "invalid fault events received" }, 51 { "nop_flts", FMD_TYPE_UINT64, "inapplicable fault events received" }, 52 { "auto_flts", FMD_TYPE_UINT64, "auto-close faults received" } 53 }; 54 55 typedef struct cma_subscriber { 56 const char *subr_class; 57 const char *subr_sname; 58 uint_t subr_svers; 59 void (*subr_func)(fmd_hdl_t *, nvlist_t *, nvlist_t *, const char *); 60 } cma_subscriber_t; 61 62 static const cma_subscriber_t cma_subrs[] = { 63 { "fault.memory.page", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 64 cma_page_retire }, 65 { "fault.memory.dimm", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 66 NULL }, 67 { "fault.memory.bank", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 68 NULL }, 69 { "fault.memory.datapath", FM_FMRI_SCHEME_MEM, FM_MEM_SCHEME_VERSION, 70 NULL }, 71 72 /* 73 * The following ultraSPARC-T1 faults do NOT retire a cpu thread, 74 * and therefore must be intercepted before 75 * the default "fault.cpu.*" dispatch to cma_cpu_retire. 76 */ 77 { "fault.cpu.ultraSPARC-T1.freg", FM_FMRI_SCHEME_CPU, 78 FM_CPU_SCHEME_VERSION, NULL }, 79 { "fault.cpu.ultraSPARC-T1.l2cachedata", FM_FMRI_SCHEME_CPU, 80 FM_CPU_SCHEME_VERSION, NULL }, 81 { "fault.cpu.ultraSPARC-T1.l2cachetag", FM_FMRI_SCHEME_CPU, 82 FM_CPU_SCHEME_VERSION, NULL }, 83 { "fault.cpu.ultraSPARC-T1.l2cachectl", FM_FMRI_SCHEME_CPU, 84 FM_CPU_SCHEME_VERSION, NULL }, 85 { "fault.cpu.ultraSPARC-T1.mau", FM_FMRI_SCHEME_CPU, 86 FM_CPU_SCHEME_VERSION, NULL }, 87 { "fault.cpu.*", FM_FMRI_SCHEME_CPU, FM_CPU_SCHEME_VERSION, 88 cma_cpu_retire }, 89 { NULL, NULL, 0, NULL } 90 }; 91 92 static const cma_subscriber_t * 93 nvl2subr(fmd_hdl_t *hdl, nvlist_t *nvl, nvlist_t **asrup) 94 { 95 const cma_subscriber_t *sp; 96 nvlist_t *asru; 97 char *scheme; 98 uint8_t version; 99 100 if (nvlist_lookup_nvlist(nvl, FM_FAULT_ASRU, &asru) != 0 || 101 nvlist_lookup_string(asru, FM_FMRI_SCHEME, &scheme) != 0 || 102 nvlist_lookup_uint8(asru, FM_VERSION, &version) != 0) { 103 cma_stats.bad_flts.fmds_value.ui64++; 104 return (NULL); 105 } 106 107 for (sp = cma_subrs; sp->subr_class != NULL; sp++) { 108 if (fmd_nvl_class_match(hdl, nvl, sp->subr_class) && 109 strcmp(scheme, sp->subr_sname) == 0 && 110 version <= sp->subr_svers) { 111 *asrup = asru; 112 return (sp); 113 } 114 } 115 116 cma_stats.nop_flts.fmds_value.ui64++; 117 return (NULL); 118 } 119 120 static void 121 cma_recv_list(fmd_hdl_t *hdl, nvlist_t *nvl) 122 { 123 char *uuid = NULL; 124 nvlist_t **nva; 125 uint_t nvc = 0; 126 int err = 0; 127 128 err |= nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid); 129 err |= nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST, 130 &nva, &nvc); 131 if (err != 0) { 132 cma_stats.bad_flts.fmds_value.ui64++; 133 return; 134 } 135 136 while (nvc-- != 0 && !fmd_case_uuclosed(hdl, uuid)) { 137 nvlist_t *nvl = *nva++; 138 const cma_subscriber_t *subr; 139 nvlist_t *asru; 140 141 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 142 continue; 143 144 if (subr->subr_func != NULL) 145 subr->subr_func(hdl, nvl, asru, uuid); 146 } 147 } 148 149 static void 150 cma_recv_one(fmd_hdl_t *hdl, nvlist_t *nvl) 151 { 152 const cma_subscriber_t *subr; 153 nvlist_t *asru; 154 155 if ((subr = nvl2subr(hdl, nvl, &asru)) == NULL) 156 return; 157 158 if (subr->subr_func != NULL) 159 subr->subr_func(hdl, nvl, asru, NULL); 160 } 161 162 /*ARGSUSED*/ 163 static void 164 cma_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) 165 { 166 fmd_hdl_debug(hdl, "received %s\n", class); 167 168 if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0) 169 cma_recv_list(hdl, nvl); 170 else 171 cma_recv_one(hdl, nvl); 172 } 173 174 /*ARGSUSED*/ 175 static void 176 cma_timeout(fmd_hdl_t *hdl, id_t id, void *arg) 177 { 178 if (id == cma.cma_page_timerid) 179 cma_page_retry(hdl); 180 } 181 182 static const fmd_hdl_ops_t fmd_ops = { 183 cma_recv, /* fmdo_recv */ 184 cma_timeout, /* fmdo_timeout */ 185 NULL, /* fmdo_close */ 186 NULL, /* fmdo_stats */ 187 NULL, /* fmdo_gc */ 188 }; 189 190 static const fmd_prop_t fmd_props[] = { 191 { "cpu_tries", FMD_TYPE_UINT32, "10" }, 192 { "cpu_delay", FMD_TYPE_TIME, "1sec" }, 193 { "cpu_offline_enable", FMD_TYPE_BOOL, "true" }, 194 { "cpu_forced_offline", FMD_TYPE_BOOL, "true" }, 195 { "cpu_blacklist_enable", FMD_TYPE_BOOL, "true" }, 196 { "page_ret_mindelay", FMD_TYPE_TIME, "1sec" }, 197 { "page_ret_maxdelay", FMD_TYPE_TIME, "5min" }, 198 { "page_retire_enable", FMD_TYPE_BOOL, "true" }, 199 { NULL, 0, NULL } 200 }; 201 202 static const fmd_hdl_info_t fmd_info = { 203 "CPU/Memory Retire Agent", CMA_VERSION, &fmd_ops, fmd_props 204 }; 205 206 void 207 _fmd_init(fmd_hdl_t *hdl) 208 { 209 hrtime_t nsec; 210 211 if (fmd_hdl_register(hdl, FMD_API_VERSION, &fmd_info) != 0) 212 return; /* invalid data in configuration file */ 213 214 fmd_hdl_subscribe(hdl, "fault.cpu.*"); 215 fmd_hdl_subscribe(hdl, "fault.memory.*"); 216 217 (void) fmd_stat_create(hdl, FMD_STAT_NOALLOC, sizeof (cma_stats) / 218 sizeof (fmd_stat_t), (fmd_stat_t *)&cma_stats); 219 220 cma.cma_cpu_tries = fmd_prop_get_int32(hdl, "cpu_tries"); 221 222 nsec = fmd_prop_get_int64(hdl, "cpu_delay"); 223 cma.cma_cpu_delay.tv_sec = nsec / NANOSEC; 224 cma.cma_cpu_delay.tv_nsec = nsec % NANOSEC; 225 226 cma.cma_page_mindelay = fmd_prop_get_int64(hdl, "page_ret_mindelay"); 227 cma.cma_page_maxdelay = fmd_prop_get_int64(hdl, "page_ret_maxdelay"); 228 229 cma.cma_cpu_dooffline = fmd_prop_get_int32(hdl, "cpu_offline_enable"); 230 cma.cma_cpu_forcedoffline = fmd_prop_get_int32(hdl, 231 "cpu_forced_offline"); 232 cma.cma_cpu_doblacklist = fmd_prop_get_int32(hdl, 233 "cpu_blacklist_enable"); 234 cma.cma_page_doretire = fmd_prop_get_int32(hdl, "page_retire_enable"); 235 236 if (cma.cma_page_maxdelay < cma.cma_page_mindelay) 237 fmd_hdl_abort(hdl, "page retirement delays conflict\n"); 238 } 239 240 void 241 _fmd_fini(fmd_hdl_t *hdl) 242 { 243 cma_page_fini(hdl); 244 } 245